https://github.com/torvalds/linux
Revision 9230a0b65b47fe6856c4468ec0175c4987e5bede authored by Dave Chinner on 20 November 2018, 06:50:08 UTC, committed by Darrick J. Wong on 21 November 2018, 18:10:53 UTC
Long saga. There have been days spent following this through dead end after dead end in multi-GB event traces. This morning, after writing a trace-cmd wrapper that enabled me to be more selective about XFS trace points, I discovered that I could get just enough essential tracepoints enabled that there was a 50:50 chance the fsx config would fail at ~115k ops. If it didn't fail at op 115547, I stopped fsx at op 115548 anyway. That gave me two traces - one where the problem manifested, and one where it didn't. After refining the traces to have the necessary information, I found that in the failing case there was a real extent in the COW fork compared to an unwritten extent in the working case. Walking back through the two traces to the point where the CWO fork extents actually diverged, I found that the bad case had an extra unwritten extent in it. This is likely because the bug it led me to had triggered multiple times in those 115k ops, leaving stray COW extents around. What I saw was a COW delalloc conversion to an unwritten extent (as they should always be through xfs_iomap_write_allocate()) resulted in a /written extent/: xfs_writepage: dev 259:0 ino 0x83 pgoff 0x17000 size 0x79a00 offset 0 length 0 xfs_iext_remove: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/2 offset 32 block 152 count 20 flag 1 caller xfs_bmap_add_extent_delay_real xfs_bmap_pre_update: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 4503599627239429 count 31 flag 0 caller xfs_bmap_add_extent_delay_real xfs_bmap_post_update: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 121 count 51 flag 0 caller xfs_bmap_add_ex Basically, Cow fork before: 0 1 32 52 +H+DDDDDDDDDDDD+UUUUUUUUUUU+ PREV RIGHT COW delalloc conversion allocates: 1 32 +uuuuuuuuuuuu+ NEW And the result according to the xfs_bmap_post_update trace was: 0 1 32 52 +H+wwwwwwwwwwwwwwwwwwwwwwww+ PREV Which is clearly wrong - it should be a merged unwritten extent, not an unwritten extent. That lead me to look at the LEFT_FILLING|RIGHT_FILLING|RIGHT_CONTIG case in xfs_bmap_add_extent_delay_real(), and sure enough, there's the bug. It takes the old delalloc extent (PREV) and adds the length of the RIGHT extent to it, takes the start block from NEW, removes the RIGHT extent and then updates PREV with the new extent. What it fails to do is update PREV.br_state. For delalloc, this is always XFS_EXT_NORM, while in this case we are converting the delayed allocation to unwritten, so it needs to be updated to XFS_EXT_UNWRITTEN. This LF|RF|RC case does not do this, and so the resultant extent is always written. And that's the bug I've been chasing for a week - a bmap btree bug, not a reflink/dedupe/copy_file_range bug, but a BMBT bug introduced with the recent in core extent tree scalability enhancements. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
1 parent 2c30717
Tip revision: 9230a0b65b47fe6856c4468ec0175c4987e5bede authored by Dave Chinner on 20 November 2018, 06:50:08 UTC
xfs: delalloc -> unwritten COW fork allocation can go wrong
xfs: delalloc -> unwritten COW fork allocation can go wrong
Tip revision: 9230a0b
bsg.c
/*
* bsg.c - block layer implementation of the sg v4 interface
*
* Copyright (C) 2004 Jens Axboe <axboe@suse.de> SUSE Labs
* Copyright (C) 2004 Peter M. Jones <pjones@redhat.com>
*
* This file is subject to the terms and conditions of the GNU General Public
* License version 2. See the file "COPYING" in the main directory of this
* archive for more details.
*
*/
#include <linux/module.h>
#include <linux/init.h>
#include <linux/file.h>
#include <linux/blkdev.h>
#include <linux/cdev.h>
#include <linux/jiffies.h>
#include <linux/percpu.h>
#include <linux/idr.h>
#include <linux/bsg.h>
#include <linux/slab.h>
#include <scsi/scsi.h>
#include <scsi/scsi_ioctl.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_driver.h>
#include <scsi/sg.h>
#define BSG_DESCRIPTION "Block layer SCSI generic (bsg) driver"
#define BSG_VERSION "0.4"
#define bsg_dbg(bd, fmt, ...) \
pr_debug("%s: " fmt, (bd)->name, ##__VA_ARGS__)
struct bsg_device {
struct request_queue *queue;
spinlock_t lock;
struct hlist_node dev_list;
refcount_t ref_count;
char name[20];
int max_queue;
};
#define BSG_DEFAULT_CMDS 64
#define BSG_MAX_DEVS 32768
static DEFINE_MUTEX(bsg_mutex);
static DEFINE_IDR(bsg_minor_idr);
#define BSG_LIST_ARRAY_SIZE 8
static struct hlist_head bsg_device_list[BSG_LIST_ARRAY_SIZE];
static struct class *bsg_class;
static int bsg_major;
static inline struct hlist_head *bsg_dev_idx_hash(int index)
{
return &bsg_device_list[index & (BSG_LIST_ARRAY_SIZE - 1)];
}
#define uptr64(val) ((void __user *)(uintptr_t)(val))
static int bsg_scsi_check_proto(struct sg_io_v4 *hdr)
{
if (hdr->protocol != BSG_PROTOCOL_SCSI ||
hdr->subprotocol != BSG_SUB_PROTOCOL_SCSI_CMD)
return -EINVAL;
return 0;
}
static int bsg_scsi_fill_hdr(struct request *rq, struct sg_io_v4 *hdr,
fmode_t mode)
{
struct scsi_request *sreq = scsi_req(rq);
sreq->cmd_len = hdr->request_len;
if (sreq->cmd_len > BLK_MAX_CDB) {
sreq->cmd = kzalloc(sreq->cmd_len, GFP_KERNEL);
if (!sreq->cmd)
return -ENOMEM;
}
if (copy_from_user(sreq->cmd, uptr64(hdr->request), sreq->cmd_len))
return -EFAULT;
if (blk_verify_command(sreq->cmd, mode))
return -EPERM;
return 0;
}
static int bsg_scsi_complete_rq(struct request *rq, struct sg_io_v4 *hdr)
{
struct scsi_request *sreq = scsi_req(rq);
int ret = 0;
/*
* fill in all the output members
*/
hdr->device_status = sreq->result & 0xff;
hdr->transport_status = host_byte(sreq->result);
hdr->driver_status = driver_byte(sreq->result);
hdr->info = 0;
if (hdr->device_status || hdr->transport_status || hdr->driver_status)
hdr->info |= SG_INFO_CHECK;
hdr->response_len = 0;
if (sreq->sense_len && hdr->response) {
int len = min_t(unsigned int, hdr->max_response_len,
sreq->sense_len);
if (copy_to_user(uptr64(hdr->response), sreq->sense, len))
ret = -EFAULT;
else
hdr->response_len = len;
}
if (rq->next_rq) {
hdr->dout_resid = sreq->resid_len;
hdr->din_resid = scsi_req(rq->next_rq)->resid_len;
} else if (rq_data_dir(rq) == READ) {
hdr->din_resid = sreq->resid_len;
} else {
hdr->dout_resid = sreq->resid_len;
}
return ret;
}
static void bsg_scsi_free_rq(struct request *rq)
{
scsi_req_free_cmd(scsi_req(rq));
}
static const struct bsg_ops bsg_scsi_ops = {
.check_proto = bsg_scsi_check_proto,
.fill_hdr = bsg_scsi_fill_hdr,
.complete_rq = bsg_scsi_complete_rq,
.free_rq = bsg_scsi_free_rq,
};
static struct request *
bsg_map_hdr(struct request_queue *q, struct sg_io_v4 *hdr, fmode_t mode)
{
struct request *rq, *next_rq = NULL;
int ret;
if (!q->bsg_dev.class_dev)
return ERR_PTR(-ENXIO);
if (hdr->guard != 'Q')
return ERR_PTR(-EINVAL);
ret = q->bsg_dev.ops->check_proto(hdr);
if (ret)
return ERR_PTR(ret);
rq = blk_get_request(q, hdr->dout_xfer_len ?
REQ_OP_SCSI_OUT : REQ_OP_SCSI_IN, 0);
if (IS_ERR(rq))
return rq;
ret = q->bsg_dev.ops->fill_hdr(rq, hdr, mode);
if (ret)
goto out;
rq->timeout = msecs_to_jiffies(hdr->timeout);
if (!rq->timeout)
rq->timeout = q->sg_timeout;
if (!rq->timeout)
rq->timeout = BLK_DEFAULT_SG_TIMEOUT;
if (rq->timeout < BLK_MIN_SG_TIMEOUT)
rq->timeout = BLK_MIN_SG_TIMEOUT;
if (hdr->dout_xfer_len && hdr->din_xfer_len) {
if (!test_bit(QUEUE_FLAG_BIDI, &q->queue_flags)) {
ret = -EOPNOTSUPP;
goto out;
}
next_rq = blk_get_request(q, REQ_OP_SCSI_IN, 0);
if (IS_ERR(next_rq)) {
ret = PTR_ERR(next_rq);
goto out;
}
rq->next_rq = next_rq;
ret = blk_rq_map_user(q, next_rq, NULL, uptr64(hdr->din_xferp),
hdr->din_xfer_len, GFP_KERNEL);
if (ret)
goto out_free_nextrq;
}
if (hdr->dout_xfer_len) {
ret = blk_rq_map_user(q, rq, NULL, uptr64(hdr->dout_xferp),
hdr->dout_xfer_len, GFP_KERNEL);
} else if (hdr->din_xfer_len) {
ret = blk_rq_map_user(q, rq, NULL, uptr64(hdr->din_xferp),
hdr->din_xfer_len, GFP_KERNEL);
}
if (ret)
goto out_unmap_nextrq;
return rq;
out_unmap_nextrq:
if (rq->next_rq)
blk_rq_unmap_user(rq->next_rq->bio);
out_free_nextrq:
if (rq->next_rq)
blk_put_request(rq->next_rq);
out:
q->bsg_dev.ops->free_rq(rq);
blk_put_request(rq);
return ERR_PTR(ret);
}
static int blk_complete_sgv4_hdr_rq(struct request *rq, struct sg_io_v4 *hdr,
struct bio *bio, struct bio *bidi_bio)
{
int ret;
ret = rq->q->bsg_dev.ops->complete_rq(rq, hdr);
if (rq->next_rq) {
blk_rq_unmap_user(bidi_bio);
blk_put_request(rq->next_rq);
}
blk_rq_unmap_user(bio);
rq->q->bsg_dev.ops->free_rq(rq);
blk_put_request(rq);
return ret;
}
static struct bsg_device *bsg_alloc_device(void)
{
struct bsg_device *bd;
bd = kzalloc(sizeof(struct bsg_device), GFP_KERNEL);
if (unlikely(!bd))
return NULL;
spin_lock_init(&bd->lock);
bd->max_queue = BSG_DEFAULT_CMDS;
INIT_HLIST_NODE(&bd->dev_list);
return bd;
}
static int bsg_put_device(struct bsg_device *bd)
{
struct request_queue *q = bd->queue;
mutex_lock(&bsg_mutex);
if (!refcount_dec_and_test(&bd->ref_count)) {
mutex_unlock(&bsg_mutex);
return 0;
}
hlist_del(&bd->dev_list);
mutex_unlock(&bsg_mutex);
bsg_dbg(bd, "tearing down\n");
/*
* close can always block
*/
kfree(bd);
blk_put_queue(q);
return 0;
}
static struct bsg_device *bsg_add_device(struct inode *inode,
struct request_queue *rq,
struct file *file)
{
struct bsg_device *bd;
unsigned char buf[32];
lockdep_assert_held(&bsg_mutex);
if (!blk_get_queue(rq))
return ERR_PTR(-ENXIO);
bd = bsg_alloc_device();
if (!bd) {
blk_put_queue(rq);
return ERR_PTR(-ENOMEM);
}
bd->queue = rq;
refcount_set(&bd->ref_count, 1);
hlist_add_head(&bd->dev_list, bsg_dev_idx_hash(iminor(inode)));
strncpy(bd->name, dev_name(rq->bsg_dev.class_dev), sizeof(bd->name) - 1);
bsg_dbg(bd, "bound to <%s>, max queue %d\n",
format_dev_t(buf, inode->i_rdev), bd->max_queue);
return bd;
}
static struct bsg_device *__bsg_get_device(int minor, struct request_queue *q)
{
struct bsg_device *bd;
lockdep_assert_held(&bsg_mutex);
hlist_for_each_entry(bd, bsg_dev_idx_hash(minor), dev_list) {
if (bd->queue == q) {
refcount_inc(&bd->ref_count);
goto found;
}
}
bd = NULL;
found:
return bd;
}
static struct bsg_device *bsg_get_device(struct inode *inode, struct file *file)
{
struct bsg_device *bd;
struct bsg_class_device *bcd;
/*
* find the class device
*/
mutex_lock(&bsg_mutex);
bcd = idr_find(&bsg_minor_idr, iminor(inode));
if (!bcd) {
bd = ERR_PTR(-ENODEV);
goto out_unlock;
}
bd = __bsg_get_device(iminor(inode), bcd->queue);
if (!bd)
bd = bsg_add_device(inode, bcd->queue, file);
out_unlock:
mutex_unlock(&bsg_mutex);
return bd;
}
static int bsg_open(struct inode *inode, struct file *file)
{
struct bsg_device *bd;
bd = bsg_get_device(inode, file);
if (IS_ERR(bd))
return PTR_ERR(bd);
file->private_data = bd;
return 0;
}
static int bsg_release(struct inode *inode, struct file *file)
{
struct bsg_device *bd = file->private_data;
file->private_data = NULL;
return bsg_put_device(bd);
}
static long bsg_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
struct bsg_device *bd = file->private_data;
int __user *uarg = (int __user *) arg;
int ret;
switch (cmd) {
/*
* our own ioctls
*/
case SG_GET_COMMAND_Q:
return put_user(bd->max_queue, uarg);
case SG_SET_COMMAND_Q: {
int queue;
if (get_user(queue, uarg))
return -EFAULT;
if (queue < 1)
return -EINVAL;
spin_lock_irq(&bd->lock);
bd->max_queue = queue;
spin_unlock_irq(&bd->lock);
return 0;
}
/*
* SCSI/sg ioctls
*/
case SG_GET_VERSION_NUM:
case SCSI_IOCTL_GET_IDLUN:
case SCSI_IOCTL_GET_BUS_NUMBER:
case SG_SET_TIMEOUT:
case SG_GET_TIMEOUT:
case SG_GET_RESERVED_SIZE:
case SG_SET_RESERVED_SIZE:
case SG_EMULATED_HOST:
case SCSI_IOCTL_SEND_COMMAND: {
void __user *uarg = (void __user *) arg;
return scsi_cmd_ioctl(bd->queue, NULL, file->f_mode, cmd, uarg);
}
case SG_IO: {
struct request *rq;
struct bio *bio, *bidi_bio = NULL;
struct sg_io_v4 hdr;
int at_head;
if (copy_from_user(&hdr, uarg, sizeof(hdr)))
return -EFAULT;
rq = bsg_map_hdr(bd->queue, &hdr, file->f_mode);
if (IS_ERR(rq))
return PTR_ERR(rq);
bio = rq->bio;
if (rq->next_rq)
bidi_bio = rq->next_rq->bio;
at_head = (0 == (hdr.flags & BSG_FLAG_Q_AT_TAIL));
blk_execute_rq(bd->queue, NULL, rq, at_head);
ret = blk_complete_sgv4_hdr_rq(rq, &hdr, bio, bidi_bio);
if (copy_to_user(uarg, &hdr, sizeof(hdr)))
return -EFAULT;
return ret;
}
default:
return -ENOTTY;
}
}
static const struct file_operations bsg_fops = {
.open = bsg_open,
.release = bsg_release,
.unlocked_ioctl = bsg_ioctl,
.owner = THIS_MODULE,
.llseek = default_llseek,
};
void bsg_unregister_queue(struct request_queue *q)
{
struct bsg_class_device *bcd = &q->bsg_dev;
if (!bcd->class_dev)
return;
mutex_lock(&bsg_mutex);
idr_remove(&bsg_minor_idr, bcd->minor);
if (q->kobj.sd)
sysfs_remove_link(&q->kobj, "bsg");
device_unregister(bcd->class_dev);
bcd->class_dev = NULL;
mutex_unlock(&bsg_mutex);
}
EXPORT_SYMBOL_GPL(bsg_unregister_queue);
int bsg_register_queue(struct request_queue *q, struct device *parent,
const char *name, const struct bsg_ops *ops)
{
struct bsg_class_device *bcd;
dev_t dev;
int ret;
struct device *class_dev = NULL;
/*
* we need a proper transport to send commands, not a stacked device
*/
if (!queue_is_rq_based(q))
return 0;
bcd = &q->bsg_dev;
memset(bcd, 0, sizeof(*bcd));
mutex_lock(&bsg_mutex);
ret = idr_alloc(&bsg_minor_idr, bcd, 0, BSG_MAX_DEVS, GFP_KERNEL);
if (ret < 0) {
if (ret == -ENOSPC) {
printk(KERN_ERR "bsg: too many bsg devices\n");
ret = -EINVAL;
}
goto unlock;
}
bcd->minor = ret;
bcd->queue = q;
bcd->ops = ops;
dev = MKDEV(bsg_major, bcd->minor);
class_dev = device_create(bsg_class, parent, dev, NULL, "%s", name);
if (IS_ERR(class_dev)) {
ret = PTR_ERR(class_dev);
goto idr_remove;
}
bcd->class_dev = class_dev;
if (q->kobj.sd) {
ret = sysfs_create_link(&q->kobj, &bcd->class_dev->kobj, "bsg");
if (ret)
goto unregister_class_dev;
}
mutex_unlock(&bsg_mutex);
return 0;
unregister_class_dev:
device_unregister(class_dev);
idr_remove:
idr_remove(&bsg_minor_idr, bcd->minor);
unlock:
mutex_unlock(&bsg_mutex);
return ret;
}
int bsg_scsi_register_queue(struct request_queue *q, struct device *parent)
{
if (!blk_queue_scsi_passthrough(q)) {
WARN_ONCE(true, "Attempt to register a non-SCSI queue\n");
return -EINVAL;
}
return bsg_register_queue(q, parent, dev_name(parent), &bsg_scsi_ops);
}
EXPORT_SYMBOL_GPL(bsg_scsi_register_queue);
static struct cdev bsg_cdev;
static char *bsg_devnode(struct device *dev, umode_t *mode)
{
return kasprintf(GFP_KERNEL, "bsg/%s", dev_name(dev));
}
static int __init bsg_init(void)
{
int ret, i;
dev_t devid;
for (i = 0; i < BSG_LIST_ARRAY_SIZE; i++)
INIT_HLIST_HEAD(&bsg_device_list[i]);
bsg_class = class_create(THIS_MODULE, "bsg");
if (IS_ERR(bsg_class))
return PTR_ERR(bsg_class);
bsg_class->devnode = bsg_devnode;
ret = alloc_chrdev_region(&devid, 0, BSG_MAX_DEVS, "bsg");
if (ret)
goto destroy_bsg_class;
bsg_major = MAJOR(devid);
cdev_init(&bsg_cdev, &bsg_fops);
ret = cdev_add(&bsg_cdev, MKDEV(bsg_major, 0), BSG_MAX_DEVS);
if (ret)
goto unregister_chrdev;
printk(KERN_INFO BSG_DESCRIPTION " version " BSG_VERSION
" loaded (major %d)\n", bsg_major);
return 0;
unregister_chrdev:
unregister_chrdev_region(MKDEV(bsg_major, 0), BSG_MAX_DEVS);
destroy_bsg_class:
class_destroy(bsg_class);
return ret;
}
MODULE_AUTHOR("Jens Axboe");
MODULE_DESCRIPTION(BSG_DESCRIPTION);
MODULE_LICENSE("GPL");
device_initcall(bsg_init);
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...